Pantograph Structure Anomaly Detection Based on Computer Vision

Pantograph is the only equipment for trains to obtain power. As the core component of the train power supply system, real-time and accurate detection of the pantograph structure state is of great significance to ensure the safety of train operation. Aiming at the shortcomings of current pantograph structure anomaly detection method, such as high cost, narrow application range and strong dependence on abnormal data, this paper designs a pantograph structure anomaly detection method based on computer vision technology, which has a wide range of application and does not rely on abnormal data. Firstly, the improved YOLOv5s object detection model is used to locate the pantograph area in the vehicle pantograph surveillance camera in real time. Then, the semantic segmentation method based on UNet is used to accurately segment the pantograph structure. Finally, the state of the pantograph structure is judged by calculating the similarity with the normal state of the pantograph structure. The experimental results show that the detection method can realize the detection of pantograph structural anomalies with high accuracy and speed.